ANEW ANALYSIS OFTHELThffiUVERB

GEORGE V AN DRIEM

1. THE LIMBU VERB

Limbu is a Kiranti language native to eastern Nepal and the western fringe of Sikkim. The author devoted the fourth chapter of his 1987 grammar of the Phedappe dialect of Limbu to a morphological analysis of the Limbu simplex verb (van Driem 1987). Simplicia are non­periphrastic finite indicative verb forms, from which the various Limbu periphrastic tense forms, gerunds, participles, adhortative and optative forms are derived. Limbu distinguishes eleven pronominal categories: first, second and third person, singular, dual and plural number, and an inclusive-exclusive distinction in the first person dual and plural. The Limbu transitive verb shows agreement with both agent (transitive subject) and patient (transitive object or beneficiary), and the transitive paradigm distinguishes 44 different forms, as shown in Tables 2 and 3. Limbu intransitive and reflexive verbs agree with the subject (intransitive or reflexive subject), and the intransitive and reflexive paradigm distinguishes eleven different forms, as shown in Table 1.

After I had completed the manuscript for the Limbu grammar in the summer of 1986, Professor Emeritus Carl Ebeling of the University of Amsterdam and Caucasologist Rieks Smeets of Leiden University suggested that alternative analyses of Limbu conjugational morphology were possible, some of which might require positing fewer slots or functional positions. Since I was eager to start my work on the Dumi language in the late summer of 1986, I did not give the matter high priority at the time. In subsequent comparative work on conjugational morphology in Kiranti and other Tibeto-Burman languages, I proposed morphological analyses for the conjugations of other languages, none of which, as it turned out, presumed as many suffixal slots as my Limbu analysis. Even the morphological analysis of the older Bahing paradigm, the transitive conjugation of which distinguished as many as 64 of the 75 theoretically possible forms, presumed fewer slots than the Limbu analysis. As time went on, my dissatisfaction with the 1987 analysis grew, and I began to assign the students of my Limbu course at Leiden University the task of revising my 1987 morphological analysis, an exercise which proved both instructive and entertaining. Recently, Ebert (1991, 1992) presented papers, which included diagrams of morphemic analyses of Limbu, Charnling and Athpare simplicia. The complete Charnling and Athpare data have not been made available, but her Limbu diagram represents an insufficient analysis. Because of the relevance of Limbu conjugational morphology to the comparative study of Tibeto-Burman verbal flexion, therefore, I felt it was high time to present a new morphological analysis of verbal agreement in the Limbu simplex.

David Bradley, ed. Papers in Southeast Asian linguistics No.l4: Tibeto-Burman languages of the Himalayas, 157-173. Pacific Linguistics, A-86, 1997.

© George van Driem 157

158 GEORGE VAN DRIEM

s

u

b

e

c

TABLE 1: LIMBU AFFIRMATIVE AND NEGATIVE INTRANSITIVE

AND REFLEXIVE SIMPLEX CONJUGATIONS IN THE NON-PRETERlTE AND PRETERlTE TENSES

(In each box the negative form is listed below the corresponding affirmative form.)

INTRANSITIVE REFLEXIVE

NPT PT NPT PT

lSG L-?E 2:'-a{] lSG 2:'-sig?E L-Sil)a{] mE-2:'- ?rn mE-J;-a{]DEil me-L-siv?rn me-L-sil)a{]DEI1

lDU.INC a-2:'-si a-2:'-Etchi lDL.INC a-2:'-nEtchi a-L-netchi an-J;-sin an-2:'-Etchin an-L-nctchin an-L-nEtchin

lDU.EXC 2:'-sige L-Etchige lDU.EXC 2:'-netchige L-nEtchige me-2:'-sigrn me-2:'-Etchigrn me-L-netchigrn me-I:-nEtchigrn

lPL.INC a-2:' a-2:'-e lPL.INC a-L-siiJ a-2:'-sil)£ an-J;-nrn an-E-rn an-2:'-sivnm an-2:'-sil)rn

lPL.EXC 2:'-ige 2:'-m?na lPL.EXC E-sivige E-siiJIJ?na me-2:'-iurn mrn-2:'-m?na me-2:'-siiJiarn mrn-L-siiJIJ?na

2SG kE-L k£-L-e 2SG k£-L-siiJ kE-J;-sil)£

krn-L-nrn krn-L-rn krn-I:-sil)nm km-L-sii}rn

2DU kE-L-si ke-L-etchi 2DU kE-L-nctchi ke-L-nctchi

krn-L-sin krn-2:'-etchin krn-L-netchin krn-L-nEtchin

2PL k£-2:'-i ke-2:'-i 2PL ke-L-sivi kE-L-sil)i krn-2:'-in krn-2:'-in km-L-siiJin krn-L-siiJin

3SG E L-e 3SG L-siv L-sil)£ mE-L-nrn me-E-rn me-L-sil)nrn me-L-sii}rn

3DU E-si 2:'-etchi 3DU 2:'-netchi L-netchi me-L-sin me-2:'-Etchin me-2:'-nctchin me-L-netchin

3PL me-2:' mE-L-E 3PL me-L-siiJ me-L-sil)£ mrn-L-nrn mrn-L-rn mrn-L-siiJnED mrn-2:'-sivrn

In the following, I shall use the words 'old' and 'new' to refer respectively to the morphemes, morpheme labels and slots of the 1987 analysis and of the new analysis proposed here. New slots are proposed, and some old slots have been abolished. Certain morpheme labels have been made more precise. Zero morphemes have been reassessed, and the problem of negation is discussed. Implications of the new analysis for the diachronic view of conjugational morphology in Kiranti and in Tibeto-Burrnan are discussed.

Abbreviations used in this paper are:

1, 2, 3 AG DU EXC INC NDU NEG NPT NSG PAT PF

first, second, third person agent of a transitive verb dual exclusive inclusive non-dual negative non-preterite non-singular or patient of a transitive verb prefixal slot

2. THE PREFIXAL CHAIN EXPANDED

A NEW ANALYSIS OF THE LIMBU VERB 159

PL PT REF s

SF SG

plural preterite reflexive subject of an intransitive or

reflexive verb sufflxal slot singular indicates the direction of a

transitive relationship verb stem

A flaw in the old analysis is that the old prefixal slot PFl could be occupied by two morphemes, viz. any combination of the first person prefix a- , the second person prefix k£­and an old third person zero morpheme 0. SlotS are functional positions in the affixal string of a verb, each of which can be occupied by a definable set of morphemes. The morphemes sharing a position in a string define the function of that position. There appears to be a general tendency for semantically related morphemes to occupy the same slot. Slots are language-specific and analysis-dependent and represent the non-random sequential ordering of morphemes in conjugated verb forms. To have more than one morpheme occupying a slot defeats the purpose of having slots in the first place and necessitates making statements about the relative position of two morphemes within a single slot, something which is not ascertainable in those cases in which one of these is a zero morph. Conversely, attempts at slot reduction may not be allowed to supersede the goal of formulating a maximally explanatory analysis.

The first person prefix a- always precedes the second person prefix k£- and therefore must be analysed as occupying an anterior slot. This prefix indicates first person in forms in which first person involvement is not indicated by some portmanteau. It now seems a bit overwrought to have posited a zero allomorph of this morpheme in forms containing the exclusive suffix -ge- -be (see Sprigg 1989). It is reasonable to say that the meaning of the exclusive suffix comprises the sense of first person involvement in addition to exclusion of the person or persons addressed. Michailovsky (1989:472) proposes to analyse the prefix a­as a first person non-singular inclusive morpheme despite its occurrence in 2~ 1 forms in which Michailovsky maintains that the opposition between inclusive and exclusive is effectively 'neutralised'. I cannot concur with this view, as the prefix clearly functions as a marker of first person, not only in 2~ 1 forms, but also in non-finite forms such as the supine.

The second person prefix kt:- indexes second person in forms in which second person involvement is not indexed by some portmanteau. The old third person zero morpheme may be abolished by a more precise labelling of two other morphemes in old prefixal slot PF2: The old non-singular agent/subject morpheme m£- - m- specifically marks the involvement of a third person non-singular agent or subject, and should therefore be relabelled as (3NSG.AG.S). Its abbreviated allomorph m- occurs between a preceding prefix and the root

TABLE 2: LIMBU AFFIRMATIVE TRANSITIVE SIMPLEX CONJUGATION IN THE NON-PRETERITE AND PRETERITE TENSES

(In each box the preterite form is listed below the corresponding non-preterite form.)

p a e n

ISG I DU.INC IDU.EXC IPL.INC IPL.EXC 2SG 2DU 2PL 3SG 3D 3PL

ISG l:-ns l:-nstchiv l:-niv l:-uv l:-uvsiv l:-ne l:-netchiv l:-niv 1:-uv 1:-unsiv

a IDU.INC a-2:-su a-1:-susi a-2:-stchu a-2:-stchusi

IDU.EXC 1:-netchige 1:-suge 1:-susige 1:-netchi.ae l:-etchuJ;e l:-etchusiJ;e

g lPL.INC a-2:-um a-2:-umsim a-2:-um a-2:-umsim

IPL.EXC 1:-nstchige l:-umbe l:-umsimbe 1:-ns(f::hiJ;e 1:-m ?na l:-m?nasi

e 2SG ks-J:-?s ks-2:-u ke-1:-usi ks-1:-an ks-2:-u ks-2:-usi

2DU ags-2: ags-2: ks-2:-su ke-2:-susi age-l: age-2:-e age-1:-e ks-1:-stchu ks-1:-stchusi

n 2PL age-2:-s ks-2:-um ks-2:-umsim ke-2:-um ke-2:-umsim

3SG 1:-?e a-1:-si 1:-sige a-.I 1:-ige ke-l: ke-1:-si ke-1:-i 1:-u 1:-usi l:-an a-2:-etchi l:-stchige a-2:-s l:-ige ks-1:-e ke-1:-stchi ke-1:-i 1:-u 1:-usi

3DU 2:-su l:-susi ms-1:-?s am-1:-si ms-1:-sige am-.I ms-1:-ige ksm-1: kem-1:-si ksm-1:-i 1:-etchu 1:-etchusi

3PL ms-1:-av am-1:-etchi ms-1:-t:tchige am-1:-t: me-1:-ige kem-1:-e k£m-1:-£tchi k£m-.I-i m£-1:-u m£-1:-usi

' -ms-2:-u me-2:-usi

I

...... 0\ 0

~ ~ ~ <: ~ t;:; ~

A NEW ANALYSIS OF THE UMBU VERB 161

of the verb, but the full form occurs between a preceding pn;fix and a following negative morpheme. The old singular agent/subject zero morph must be retained but more precisely labelled as the specific marker of a third person singular agent or subject (3SG.AG.S). These considerations necessitate positing more prefixal slots for person in the Limbu verb than in the old analysis, that is a first person slot (PFl), a second person slot (PF2) and a third person slot (PF3), replacing older slots PFl and PF2 ..

LIMBU PERSON AND NUMBER AGREE~NT PREFIXES

PFl PF2 PF3

me--m-a- ke- 3NSG.AG.S

1 2 f)

3SG.AG.S

3. THE PROBLEM OF NEGATION

An analysis of negation in Limbu simplicia would have to account for the fact that the prefix men- occurs in negative preterite forms with a first person singular or first person plural exclusive agent or subject and the lack of an overt negative suffix in such forms. In all other negative simplicia, there are at least two negative and no more than three overt negative affixes, of which one is a prefix.

ANALYSIS A ANALYSISB ANALYSIS C

PF4 PF5 PF4 PF4 me--n-

me- n- NEG me- - n- - men-NEG NEG NEG

men-lSG.AG.S/lPL.EXC.AG.S/PT.NEG

Analysis B posits a distinct negative morpheme men- in forms with a first person singular or first person plural exclusive agent or subject, bearing the clumsy morpheme label (lSG.AG.SilPL.EXC.AG.S/PT.NEG). The sequence men- in such forms is homophonous with the morpheme sequence me-+ n- (3NSG.AG.S + NEG) in negative forms with a non­singular third person agent or subject, although this does not lead to homophony anywhere in the paradigm. Analysis A, on the other hand, would explain the sequence men- in negative preterite forms with a first person singular or first person plural exclusive agent or subject as the eo-occurrence of two negative morphemes in adjacent slots, viz. me- + n- (NEG + NEG). An advantage of analysis A is that this eo-occurrence could be seen as a motivated phenomenon, in keeping with the regularity consistently observed elsewhere throughout the simplex paradigm that all negative forms are marked by at least two negative morphemes. In all other forms, one negative prefix is consistently seen to occur with one or two negative suffixes. The occurrence of two prefixed negative morphemes in negative preterite forms with a first person singular or first person plural exclusive agent or subject can be interpreted as being attributable to the fact that the negative suffix is 'blocked' by the negative preterite

ISG

a IDU.INC

IDU.EXC

g IPL.INC

IPL.EXC

e 2SG

2DU

n 2PL

3SG

3DU

3PL

TABLE 3: LIMBU NEGATIVE TRANSITIVE SIMPLEX CONJUGATION IN THE NON-PRETERITE AND PRETERITE TENSES

(In each box the preterite form is listed below the corresponding non-preterite form.)

p a e n

ISO IDU.INC IDU.EXC IPL.INC IPL.EXC 2SG 2DU 2PL 3SG

me-I-nen me-I-netchivnen me-I-nivnen me-I-?en me-I-nen me-I-netchinnen me-I-ninnen men-I-ban

an-I-sun an-I-etchun

me-I-netchigen me-I-sugen me-I-netchigen me-I-etchugen

an-I-umnen an-I-umnen

me-I-netchigen me-I-umben me-I-netchigen men-I-m7na

ken-I-?en - ken-I-un ken-I-avnen ken-I-un

agen-I-nen agen-I-nen ken-I-sun agen-I-nen agen-I-en agen-I-en ken-I-etchun agen-I-en ken-I-umnen

ken-I-umnen me-I-?en an-I-sin me-I-sigen an-I-nen me-I-igen ken-I-nen ken-I-sin ken-I-in me-I-un me-I-~vnen an-I-Eichin me-I-etchigen an-I-m me-I-igen ken-I-en ken-I-etchin ken-I-in me-I-un

me-I-sun men-I-?en amen-I-sin men-I-sigen amen-I-nen men-I-igen kemen-I-nen kemen-I-sin kemen-I-in me-I-etchun men-I-avnen amen-I-Eichin men-I-etchigen amen-I-en men-I-igen kern m-I-en kemen-I-Eichin kemen-I-in men-I-un

---- L__ __ ···-- - --men-I-un

3DU 3PL

me-I-enchin men-I-bansin an-I-sunchin an-I-etchunchin me-I-susigen me-I-etchusixen · an-I-umsimnen an-I-umsimnen me-I-umsimben men-I-m7nasi ken-I-unchin ken-I-unchin ken-I-sunchin .ken-I-etchunchin ken-I-umsimnen ken-I-umsimnen me-I-unchin me-I-unchin me-I-sunchin me-I-etchunchin men-I-unchin men-I-unchin

,_. 0\ N

~ ~ ~ § ~ t>:l ~

A NEW ANALYSIS OF THE LIMBU VERB 163

first person singular morpheme -parJ or the preterite first person plural exclusive agent/subject morpheme -m ?na, although it is not clear what factor could be responsible for blocking the slot of the negative suffix.

A question arising from analysis A is that of allomorphy. In the original analysis, the negative prefix is analysed as ms- - n-, whereby the allomorph ms- occurs whenever the negative prefix is the first overt morpheme in a verb, and the allomorph n- occurs when it is not. In analysis A, allomorphy could be dispensed with, but there seems to be no other reason for analysing the negative prefix in ms-hip-?s-n 'I shan't hit him' as being an altogether different morpheme from that in k£-Q-hipt-u-n 'you won't hit him'. Likewise, it seems neither motivated nor plaU'Sible to assign the negative morpheme m£- of prefixal slot PF4, assumed in analysis A, to some slot anterior to our current PFl which would be 'blocked' by the occurrence of one of the overt prefixes a- (1), k£- (2) or m£-- m­(3NSG.AG.S). Rather, it is simpler to assume two formally identical negative prefixes NEG 1 and NEGz, exhibiting the same pattern of allomorphy m£- - n- described in the original analysis, although the second negative morpheme only actually occurs in the form of its allomorph n- because it invariably eo-occurs with the first negative morpheme. The occurrence of the second negative prefix in prefixal slot PF5 is taken to be conditioned by the lack of a second negative morpheme in the suffixal string of the verb. This situation occurs only in negative preterite forms with a first person singular or first person plural exclusive agent or subject where the negative suffix is 'blocked' by the negative preterite first person singular suffix -parJ or the preterite first person plural exclusive agent/subject suffix -m?na.

These competing analyses cannot be meaningfully assessed without consideration of the suffixal negative morphemes. With the exception of preterite forms with a first person singular or first person plural exclusive agent or subject, simplicia are negated by simultaneous prefixation and suffixation of negative morphemes. The obligatory prefixed morpheme was termed the first negative morpheme, and the suffix the second negative morpheme. The second negative morpheme -n£n - -n (NEGz) is located in the last position in the suffixal string, which, in the new analysis, is suffixal slot SFlO. Negated non-preterite 1SG-'73NSG forms also obligatorily take a third negative morpheme, which is a suffix -n (NEG3) located in new suffixal slot SF6, for example m£-ni-7£-Q-chi-n 'I don't see them'. The third negative morpheme also occurs facultatively in 1DU.INC~3NSG, 2SG~3NSG, 2DU~3NSG, 3SG-'?3NSG and 3DU~3NSG forms. Because the negative prefixal sequence men- contains two nasal segments and because this sequence occurs only in those negative simplicia in which no negative suffix occurs, viz. preterite forms with a first person singular or first person plural exclusive agent or subject, it appeared intuitively satisfying to assume, as I did in the original analysis, that a Limbu simplex is negated by at least two negative morphemes. However, an analysis of the sequence msn- as both allomorphs of the negative prefix m£- - n- eo-occurring in succession within a single slot is unsatisfactory because two morphemes ought not to occur in a single slot. Alternatively, the sequence men­may be analysed as a special allomorph of this negative prefix occurring in negative preterite forms with a first person singular or first person plural exclusive agent or subject. This is analysis C.

Finally, it would appear to be more expedient to abandon the unsophisticated idea of several negative morphemes in Limbu simplicia and to adopt the concept of a single 'discontinuous morpheme' for the negative in simplicia, which is how I analysed the Limbu active participle k£-... -pa, Limbu negative active participle men-... -mna and Limbu negative

164 GEORGE VAN DRIEM

perfect gerund mEn-... - ?e:, although it somehow did not occur to me to analyse the negative in simplex verbs in these terms. I have recently adopted Hagege's term 'simulfix' for this phenomenon, in particular with respect to the non-singular second person actant number simulfix -a ... -ni in Lohorung (van Driem, forthcoming). The older Kiranti active participial ending kh;J- ... -kpa reflected in the petrified Dumi forms khgnikpa (ni-ni 'be good') 'good, comely, beautiful' and khgyhkpa (i:-ni 'be bad') 'horrid, ugly, foul, bad' is also such a discontinuous morpheme or simulfix (van Driem 1993b). A simulfix consists of two or more elements which are affixed simultaneously at different locations in the affixal string. Hagege (1986:26) also considered the neologisms diffixe, ambifixe and transfixe for this phenomenon, which he described for the Austronesian language Palau spoken in the archipelago of the same name in the Pacific Ocean between 134°05'E and 134°45'E and between 6°55'N and 8°15'N, but he prefers and adopts the term simulfixe. The case for analysing negation in Limbu simplicia as a case of simulfixation is compelling. This analysis would entail positing a single polylocational negative morpheme consisting of minimally one and maximally three elements in a well-defined pattern of paradigmatic distribution. The labels (NEG1), (NEG2) and (NEG3) will be retained to indicate the elements of the Limbu negative simulfix. The analysis proposed therefore superficially resembles the old analysis and analysis C above, although it is substantially different from either of these. The new analysis posits a single negative simulfix with one to three affixal manifestations, the occurrence of which is paradigmatically determined and one of which is always a prefix. The new model for the prefixal string of the Limbu simplex should therefore look like this:

THE LIMBU PREFIXAL STRING

PF1 PF2 PF3 PF4

m£-- m-

a- k£- 3NSG.AG.S m£- - n- - men-1 2 NEG1

0 3SG.AG.S

4. THE REFLEXIVE AND ALLOMORPHY

The old reflexive morpheme -si.g - -n£- -n was interpreted as having an allomorph -n£­-n in dual forms and an allomorph -si.g in singular and plural forms. In reflexive forms, the dual reflexive allomorphs -n£- -n always occur as part of the sequence -n£tchi, whereby the element -tchi is identified as the allomorph -tchi of the dual patient/subject morpheme -si­-tchi in suffixal slot SF3. The allomorph -tchi occurs regularly following the preterite morpheme-£, the dual reflexive allomorph -n£- -nand the 1----?2 portmanteau -n£- -n. The dual reflexive preterite sequence -n-£-tchi is homophonous with the corresponding non­preterite sequence -n£-0-tchi. In fact, consistently throughout the paradigm, cases of preterite/non-preterite homophony are satisfactorily accounted for by adjacency of a vocalic morpheme to the preterite morpheme -£, which elides. The dual reflexive allomorphs -n£ --n are homophonous with the 1----?2 portmanteau -n£- -n, which occurs in the 1SG-?2SG ending -nE, the 1SG-?2DU ending -n£tchi.g, the 1SG-?2PL ending -ni.g and the 1NSG-?2 ending -nEtchige. These endings also exhibit preterite/non-preterite homophony, and attempts

----------

A NEW ANALYSIS OF THE LIMBU VERB 165

to segment sequences like the dual reflexive ending -netchi as -net+ -chi have not proved to be illuminating.

There seems to be no semantic common denominator on the basis of which the dual reflexive allomorphs -ne- -nand the l-72 portmanteau -ne:- -n could be analysed as a single morpheme, and so they were analysed as distinct entities in the 1987 analysis. The reflexive allomorph -ne - -n is restricted to dual reflexive forms, and the portmanteau -ne --n indexes the transitive relationship between a first person agent and a second person patient. Subsequent comparative study seems to have vindicated this synchronic analysis. Whereas the dual reflexive allomorphs -ne - -n appear to be cognate with the Tibeto-Burman root for 'two' *g-ni-s, also reflected in the Limbu regular numeral netchi 'two' (cf. sumsi 'three', lisi 'four', nasi 'five', all with the non-singular or 'generalised dual' suffix -si) and the collective numeral nepphu 'two' (cf. sumbhu 'three', libhu 'four', with the collective suffix -phu), the l-72 portmanteau -ne- -n has cognates in the Dumi lSG-72 suffix -n, the Hayu lSG-72 suffix -no, the Kulung lSG-72 suffix -an, the Thulung lSG-72 suffix -nini, the Bahing lSG-72 suffix -na and the Jinghpaw l-72 suffix -nte?l- -te? 1, all pointing to the existence of some lSG-72 proto-morpheme at the Proto Tibeto-Burman level, which has tentatively been reconstructed as *-nya. '

ANALYSISD

SFl

-si_g - -ne - -n

REF

-ne- -n

l-72

ANALYSISE

SFl

-si_g

NDU.REF

-ne- -n

DU.REF

0

-ne- -n

The reflexive morph -ne - -n may be analysed as a specific dual reflexive morpheme contrasting with a reflexive morpheme with non-dual meaning -si_g, as in analysis E. Alternatively, the reflexive morph -ne- -n may be analysed as an allomorph of the reflexive morpheme in dual forms alongside an allomorph -si_g indexing the reflexive in non-dual forms restricted to dual forms, as in analysis D. The latter analysis in terms of allomorphy is more satisfactory because it assumes a single morpheme -si_g - -ne - -n with a single meaning and a well-defmed pattern of allomorphy. This reflexive morpheme does not index actant number because this meaning is indexed by other morphemes in the affixal string. Analysis E may reflect the distinct provenance of the two morphs, although such a diachronic criterion, even if correct, need not be of decisive synchronic importance.

5. THE DUAL MORPHEJ\ffi

The second suffixal slot SF2 is the tense slot. Non-preterite time is marked by zero, and preterite time is marked by the suffix -e- 0 with a zero allomorph before another vowel.

166 GEORGE VAN DRIEM

The non-preterite zero suffix does not occur in forms in which tense is indexed by the non­preterite frrst person singular patient/subject morpheme -?e.

The old dual agent morpheme -s - -tch, the sole occupant of old suffixal slot SF3, immediately precedes and invariably eo-occurs with the third person patient morpheme -u. Old sufflxal slot SF3 could be eliminated by analysing the sequence -s + -u, or -tch - -u after the preterite morpheme -e, as a DU--+3 portmanteau morpheme -su- -tchu, indexing the transitive relationship between a dual agent and a third person patient and occupying the same position in the suffixal string as the third person patient morpheme -u. The sole advantage to approach G, however, is that it would eliminate a slot from the analysis. The drawback to analysis G is that the new portmanteau morpheme is too easily segmented into its two component parts identified in the original analysis F. The element 1-u! in the proposed DU~3 portmanteau -su - -tchu is clearly the third person patient morpheme -u, and the element -s --tch clearly indexes dual number of agent. Analysis F, therefore, does not just represent a diachronic dissection of some fused morpheme.

ANANLYSISF ANALYSIS G

SF2 SF3 SF4 SF2 SF3

-u -e- 0 -u -e- 0 3PAT

PT 3PAT PT

-s- -tch -su- -tchu DU.AG DU-?3

0 -si- -tchi 0 NPT DU.PAT.S NPT -si- -tchi

DU.PAT.S

The old dual agent morpheme -s- -tch indexes dual number of the agent in lDU.EXC-?3, lDU.EXC-?3, 2DU-?3 and 3DU-?3 forms. The old dual patient/subject morpheme -si­-tchi indexes dual number of patient in 1SG-?2DU, 3-?2DU, 3--+lDU.INC and 3-?lDU.EXC forms, and dual number of subject in all intransitive dual forms. The old dual agent morpheme -s - -tch and the old dual patient/subject morpheme -si - -tchi can be analysed as a single entity, viz. a dual morpheme -si in suffixal slot SF3, with the regular allomorphs -s - -tch before the third person patient morpheme -u, and the regular allomorphs -tch- -tchi after the preterite morpheme -e- 0, the 1-?2 portmanteau -ne- -nand the reflexive suffix -siiJ- -n£- -n. This is analysis H.

ANALYSIS H

SF2 SF3 SF4

·£- 0 PT -si- -s- -u

-tchi- -tch 3PAT

0 DU

NPT

A NEW ANALYSIS OF THE LIMBU VERB 167

6. FIRST PERSON SINGULAR MORPHEMES

The morpheme -paiJ remains puzzling, as was explained in my Limbu grammar. In modem Phedappe Limbu, the suffix -paiJ is obligatory in negative preterite lSG-73 forms, for example mc;-bi:-baiJ-si-1] 'I didn't give it to them' (1SG-73NSG.PT)- cf. mc;-bi:-'?c;-n­chi-n 'I shan't give it to them' (1SG-73NSG.NPT), but the suffix -paiJ is optional in intransitive negative preterite forms with a first person singular subject. This leads to doublets such as ya?-mc;n-Ia:k-paiJ 'I did not partake in the rice harvest dance' and ya?-mc;­ra:kt-ai]-n£n with the same basic meaning. Where such doublets exist, difference in form expresses difference in meaning, and I have recently come to understand the semantic distinction between such negative preterite forms with a first singular subject. Moreover, the fact that the suffix -paiJ is optional in intransitive forms but obligatory in transitive forms leads to the conjecture that this morpheme was originally limited to the transitive paradigm and was absorbed into the intransitive paradigm secondarily. The difference in meaning between negative preterite doublets with a first singular subject strongly supports this conjecture.

Negative preterite first singular intransiti~e forms in -paiJ co-exist with the 'regular' intransitive forms. The intransitive forms kcraiJ 'I have arrived' and khi:'?raiJ 'I am (i.e. have become) tired' 1 are negated either as m£g£rai]n£n 'I haven't arrived' and mc;khi:?raiJnen 'I am not tired' or as meJ]geppaiJ and m£1]khi:ppaiJ, respectively. The difference in meaning, as explained by my old friend Narayal). Prasad Pailyarigu of Tamphula village (pers. comm., Kathmandu, June 1993) is that the 'regular' first singular intransitive affixation in m£-2:-81]­nc;n is simply the negative alternative to its affirmative preterite counterpart in -aiJ, for example KeraiJJ]i: m£g£rai]n£nni:? 'Have I arrived or haven't I?'. The use of the suffix -p81] in negative intransitive forms, however, adds a semantic dimension to the utterance. The added implication is that something has not happened that was supposed to happen or was expected to happen. For example, as an answer to a question one can either respond with m£dhai]ai]n£n 'I didn't come up', the regular intransitive negative of thai]aiJ 'I have come up', or with the more abrupt mc;ndhaiJb81J. The latter form with the originally transitive suffix -p81] can be construed as rude because the implication is: 'I was supposed to have come up, but I just didn't'. The regular intransitive form m£dh81JaiJnen on the other hand is a simple negation and is therefore more polite, that is 'I didn't come up' or, by implication, 'I couldn't come up' .2 It appears that the meaning of the forms in -paiJ has something to do with the greater volitionality or agency of this originally transitive suffix. In combination with the lexical meaning of certain verbs the form in -paiJ is the more obvious choice, for example mc;nchibaiJ 'I'm not dead, I didn't die', suitable when the speaker has just survived a life­threatening situation or when the speaker, presumed dead, turns out not to have died. This is why the form mc;syai)n£n 'I'm not dead' is an odd, albeit acceptable, utterance in modem Limbu, for it is as if the speaker is making the remark out of the blue, and, in fact, there are reportedly few conceivable situations in which mc;syaiJnen would be a relevant utterance.

2

The stems of the Limbu verb khi:pma 7 'to get tired' are given incorrectly in the glossary of the Limbu grammar. They should be khi:?r- khi:?. Nep. 'aunu sakina 'jasto alikati bhiiv huncha.

168 GEORGE VAN DRIEM

Historically, then, the label of the morpheme -pan would have to have been (1SG~3.PT), whereas it has acquired the modern meaning (lSG.AG.S.PT) whilst retaining some of its original sense of agency or deliberacy associated with transitivity. This enables us to identify the element /p/ in -pan, originally (1SG-73.PT), with the Bahing anticipatory 2/3SG~3 copy morpheme -p, which occurs in the preterite forms of open-stem verbs before the Bahing preterite tense morpheme -ta. This is the first cognate element to be identified in a Kiranti language for any of the Bahing anticipatory copy phenomena. The original meaning of the suffix -pan (1SG-73.PT) explains the special sense in which it entered the intransitive paradigm, for example me-dhan-an-nen 'I didn't come up' versus men-dhan-ban 'I didn't do it, viz. come up', yet this explanation is diachronic. Synchronically, the preterite first person singular agent/subject portmanteau -pan could then best be reassigned to new sufftxal slot SF4, which groups this originally agentive afftx together with other agent morphemes in a single slot and may more faithfully reflect historical reality.

Analysis I, however, assumes an alternative approach to the four first person singular morphemes identified in the old analysis, viz. the non-preterite first person singular patient/subject morpheme - ?e, the preterite first person singular patient/subject suffix -an, the

f

first person singular agent morpheme -.u and the preterite first person singular agent/subject portmanteau -pan. A tenseless first person singular morpheme -an - -.u is posited, which exhibits the regular allophone -{}after the third person patient suffix -u. Conversely, the regular zero allomorph of the preterite sufftx -e- (2> occurs before the vowel-initial suffix -an- -.u (lSG). This analysis unites the old suffixes -an (lSG.PAT.S.PT) and-{} (lSG.AG). Similarly, the portmanteau -pan can be segmented into an element -p and the newly posited first person singular morpheme -an- -.u. The element -p may be identified as a preterite third person patient morpheme, which would tally with its semantically motivated occurrence in negated first person intransitive forms, whereby third person patient agreement is with the logical proposition denoted by the verb and not enacted by the first person singular agent, as in the example men-dhan-ban 'I didn't do it, viz. come up'.

Historically, the preterite third person patient sufftx -p reflects the labial character of the Proto Kiranti third person patient morpheme *-u and the plosive character of the initial /t/ of the Proto Kiranti preterite sufftx *-te. It is unclear what historical relationship exists between the Limbu preterite third person patient morpheme -p and the Bahing anticipatory 2SG/3SG~3 morpheme -p, but the occurrence of the several Bahing anticipatory copy morphemes is likewise connected with preterite tense, indexed by the suffix -ta in Bahing.

The old non-preterite first person singular patient/subject suffix -?e is relabelled as a non­preterite first person singular morpheme, rather than as a patient/subject morpheme, in accordance with its occurrence in negative lSG--73 forms. In the old analysis, the occurrence of the suffix in these forms was attributed to the reduced transitivity of a negated situation. Now, a zero morph of the third person patient suffix may be assumed to occur in non­preterite negative 1SG~3 forms before the non-preterite first person singular sufftx -?e (lSG.NPT) and in preterite lPL.EXC--73 forms before the preterite first person plural exclusive agent/subject portmanteau -m ?na.

A NEW ANALYSIS OF THE LIMBU VERB 169

ANALYSIS!

SF3 SF4 SF5

-u- 0 -?e

-si- -s- 3PAT lSG.NPT -tchi- -tch

DU -p -81] - -1)

3PAT.PT ISG

7. THE SUFFIXAL STRING COMPACTED

The old non-singular agent number morpheme -tchi in old suffixal slot SF7 occurs only in the lNSG--72 form and was posited to account for the element -tchi in the lNSG--72 ending -netchige. The fact that the l--72 portmanteau -ne- -n can be semantically and formally identified within the ending argues against defining the sequence -netchi as a lNSG-72 portmanteau eo-occurring with the exclusive suffix -ge- -be. Moreover, positing such a portmanteau would necessitate adopting some independent ad hoc way of accounting for the otherwise regular preterite/non-preterite homophony in the lNSG-72 form. The transitive paradigm does not distinguish a dual from a plural third person patient, and non-singular number of a third person patient is indexed by the third person non-singular patient morpheme -si in old suffixal slot SF8. However, there are no formal or positional reasons not to identify the old non-singular agent morpheme -tchi in lNSG.EXC-72 forms with the old third person non-singular patient morpheme -si. A single non-singular number morpheme -si - -tchi can therefore be posited, with the regular allomorph -tchi after either the preterite tense suffix -e or the l-72 portmanteau -ne. The old singular patient number zero morph 0 should likewise be analysed as a singular number morpheme. This redefined zero morph not only indexes singular number of a third person patient but also of a second person patient or subject.

The plural agent suffix -m must be assigned to new suffixal slot SF5 because the suffix follows the third person patient morpheme -u and precedes the non-singular morpheme -si - -tchi in suffixal slot SF7. The portmanteau -m ?na, which indexes a first person plural exclusive subject or agent in preterite time, must also be assigned to a position in the suffixal string before the non-singular morpheme -si - -tchi in suffixal slot SF7. The portmanteau -m ?na can best be assigned to the same suffixal position as the plural agent suffix -m to which it appears to be semantically and formally related.

The plural patient/subject morpheme -i - 0 indexes plural number of a first or second person patient in 3-71PL.INC, 3-71PL.EXC, 1SG-72PL and 3--72PL forms, and plural number of first or second person subject in intransitive forms. The zero allomorph of this morpheme occurs in indicative forms with a first person plural inclusive patient (viz. 3--71PL.INC forms) and in intransitive forms with a first person plural inclusive subject. However, the full allomorph -i of the morpheme occurs in intransitive adhortative forms with a first person plural inclusive subject. Attempts to identify the patient/subject morpheme -i- 0 with the element /-if in the dual morpheme -si- -s- -tchi - -tch are not productive. The suffix -i- 0 must be assigned to a position following the tense slot SF2 and before the copy morpheme slot, new suffixal slot SF8. Semantic arguments could be

170 GEORGE VAN DRIEM

advanced for assigning the morpheme -i- @ to either suffixal slot SF3 or to suffixal slot SF5, but in view of the heterogeneous functions of suffixal slot SF5, it seems reasonable to assign the plural patient/subject suffix -i-@ to this position.

A zero morph @ was posited in the old analysis to index singular number of a first person agent in 1SG~2SG forms and singular number of a second person agent. This zero morph is retained in the new analysis and assigned to new suffixil slot SF5. Elsewhere in the paradigm, singular number of a first person actant is indexed_independently by the first person singular morpheme -an - -g or the non-preterite first person singular morpheme -?e. Singular number of a third person agent or subject is indicated by a zero morph @ in new prefixal slot PF3. Singular number of a third person patient, second person subject and second person patient is indicated by the singular zero morph @ in new suffixal slot SF7. The new analysis therefore posits three zero morphs indexing singular number, whereas only two singular zero morphs would, in principle, be necessary to account for actant number marking in, for example, 1SG--72SG and 2SG--73SG forms. However, the distribution of the three zero morphs (viz. @ (3SG.Ag.S) in PF3, @ (SG.AG) in SF5 and @ (SG) in SF7) more accurately reflects the split ergative structure of the Limbu paradigm, whereby patient and subject roles are opposed to agent role in the fir'st and second person, and agent and subject roles are opposed to patient role in the third person and, curiously, in the preterite of first person plural exclusive forms. The posited pattern of singular number zero morphs relies less upon what Michailovsky (1989:472) calls "choix arbitraires" in the analysis of the Limbu verb than would positing two zero morphs to account for all forms, but follows instead from symmetries inherent to the Limbu afftxal agreement pattern.

SFl

-sig - -ne - -n

REF

-ne- -n

1~2

INITIAL PORTION OF THE LIMBU SUFFIXAL STRING

(suffixal slots SFl to SF5)

SF2 SF3 SF4

-e-@ -u-@

PT -si- -s - 3PAT -tchi- -tch

0 DU -p

NPT 3PAT.PT

SF5 -?e

lSG.NPT

-81]- -g

1SG

-m PL.AG

-m?na

lPL.EXC.AG.S.PT

@

SG.AG

Because, in the new analysis, the plural agent morpheme -m and the preterite first person plural agent/subject portmanteau -m?na have been assigned to suffixal slot SF5, and because the old non-singular agent morpheme -tchi has been eliminated from the analysis as an

A NEW ANALYSIS OF THE LIMBU VERB 171

independent morpheme, old suffixal slot SF7 has been abolished. The copy slot, new suffixal slot SF8, now contains copies only of morphemes occurring in new suffixal slot SF5 rather than copies of morphemes occurring in either of two suffixal slots, as in the old analysis.

The remaining suffixal slots are unproblematic. The position of the third element of the negative simulfix -n is posited as immediately preceding the slot containing the non-singular number morpheme -si, with which it invariably eo-occurs. The first person singular copy morpheme -IJ and the plural agent copy morpheme -m occur together in suffixal slot SF8. Whereas the plural agent copy morpheme -m echoes the homophonous agentive morpheme in suffixal slot SF8, the first person singular copy morpheme -IJ reiterates the notion of first singular agent indexed by either the first singular morpheme -IJ in suffixal slot SF5 or the l---72 portmanteau -n£- -n in suffixal slot SFl. In suffixal slot SF9, the exclusive suffix -ge - -be indexes first person involvement and the exclusion of second person, and the inclusion of second person in first person forms is indexed by a zero morph 0.

SF6

FINAL PORTION OF THE LIMBU SUFFIXAL STRING

(suffixal slots SF6 to SFlO)

SF7 SF8 SF9

-si -IJ -ge- -be NSG lSG EXC

0 -m 0 SG PL.AG INC

8. CONCLUSIONS

SFlO

-n£n- -n NEG2

The model of Limbu conjugational morphology has been refined. Both the number of suffixes and the number of suffixal slots have been reduced. Conversely, the number of prefixal positions identified in the analysis has increased. Morphemes have been redefined or more accurately formulated. One zero morpheme has been eliminated, and the distribution of zero morphs indexing singular actant number appears to more satisfactorily reflect the psychological reality of Limbu conjugational morphology. Negation in the Limbu simplex has been analysed as a discontinuous morpheme, for which we adopt Hagege's term 'simulfix'.

The choice between the two alternative negative preterite first person singular simplicia in the intransitive paradigm has been shown to be semantically motivated. The difference in meaning enables us to retrace the diachronic development which led to the existence of alternative negative preterite first person singular forms in the intransitive paradigm.

The number of dual morphemes has been reduced from two to one, and the number of generalised dual morphemes, that is non-singular morphemes derived from older dual *-si, has likewise been reduced from two to one. The dual morpheme occurs amidst the initial suffixes, and the generalised dual morpheme occurs in the final portion of the string, so that the expansion of the original dual meaning to encompass the notion of non-singularity is a

172 GEORGE VAN DRIEM

phenomenon associated with the final portion of the suffixal string. This corroborates the idea that this 'generalisation' of the meaning was a secondary development observed in more recent accretions.

There are two diachronic implications of this new analysis of the Limbu verb. The first of these is that it is more plausible to posit a single dual morpheme in the initial portion of the suffixal string of the Proto Kiranti verb, as in the first and second models developed for the common ancestral verbal agreement system (viz. van Driem 1990, 1992). However, this protomorpheme *-ci would have to be glossed simply as a duil.l morpheme, rather than a morpheme-indexing dual number of specifically first and second person actants. The two homophonous dual morphemes posited in the third model of the Proto Kiranti verb (viz. van Driem 1991) do not reflect distinct entities in Proto Kiranti, but subsequent developments in Bahing and Limbu resulting from the split ergative nature underlying Kiranti conjugational morphology, whereby first and second person actants are encoded in the verb according to an ergative pattern, and third person actants are encoded according to an accusative pattern.

The second diachronic implication of this new analysis is the greater verisimilitude of the simpler, first model of the Proto Kiranti verb in tefJD.S of the way it posits a single first person singular morpheme *-g in the same suffixal position as the third person patient protomorpheme *-u, that is immediately after the position of the dual protomorpheme *-ci. Whereas the Hayu and Limbu data reflect an old sequence *-u-g, Thulung and Dumi reflect an older element order *-g-u, and the conjugations of some Kiranti languages provide no indication for the relative position of these protomorphemes. In the second and third models of the Proto Kiranti verb, two additional tensed first person singular protomorphemes were posited in a position immediately anterior to the dual protomorpheme *-ci, viz. the preterite first person singular protomorpheme *-a.v and its non-preterite counterpart *-ga. These too seem to be artificial constructs not corresponding to distinct entities in the Proto Kiranti verbal agreement system, but reflecting subsequent developments in some Kiranti languages whereby a first person singular protomorpheme, perhaps best reconstructed as *-ga - -g, when immediately preceded by the preterite tense protomorpheme *-te, gave rise in some languages to a reanalysis of adjacent segments in terms of a preterite tense first person singular morpheme *-a.v.

The third model of the Proto Kiranti verb (van Driem 1991:354) should therefore be modified as follows: (1) The tensed first person singular morphemes *-a.v (lSG.PT) and *-ga (lSG.NPT) may be eliminated, and (2) instead of the two dual morphemes preceding the third person patient protomorpheme *-u, a single dual protomorpheme *-ci should be posited at this location. Data will have to be continuously reassessed if we are to trace the evolution ofTibeto-Burman conjugational systems and to discern just which phenomena are regional developments. In Tibeto-Burman languages beyond the Kirant, there is little evidence for an older element order *-u-g (3PAT-1SG), which is reflected in the Limbu and the Hayu verb, and the many reflexes of the Proto Tibeto-Burman first person singular morpheme *-g - -a.v outside of the Himalayas suggest that this morpheme occupied an anterior position following the verb in the Proto Tibeto-Burman verbal agreement system. Yet it is still too early to say just which first person singular agreement phenomena in Kiranti are the results of secondary developments. A Tibeto-Burman dual protomorpheme *-si in a position anterior to the third person patient protomorpheme *-u is well reflected outside the Himalayas, yet the hypothetical Tibeto-Burman dual patient protomorpheme *-si in the final portion of the suffixal string was posited primarily on the basis of Kiranti data (van Driem

A NEW ANALYSIS OF THE LIMBU VERB 173

1993a). This, together with the fact that the Kiranti evidence itself suggests that the posterior reflex is a later accretion, casts doubt on the existence of the latter suffix at the Tibeto­Burman level and implies that this detail in the model for Proto Tibeto-Burman verbal agreement should be revised accordingly.

REFERENCES

Driem, George van, 1987, A grammar of Limbu. Berlin: Mouton. 1990, An exploration of Proto-Kiranti verbal morphology. Acta Linguistica Hafniensia

22:27-48. 1991, Bahing and the Proto-Kiranti verb. Bulletin of the School of Oriental and African

Studies 54/2:336-356. 1992, Le proto-kiranti revisite, morphologie verbale du lohorung. Acta Linguistica

Hafniensia 24:33-75. 1993a, The Proto-Tibeto-Burman verbal agreement system. Bulletin ofthe School of Oriental

and African Studies 56/2:292-334. 1993b, A grammar of Dumi. Berlin: Mouton de Gruyter. forthcoming, A grammar of Lohorung.

Ebert, Karen, 1991, Inverse and pseudoinverse prefixes in Kiranti languages: evidence from Belhare, Athpare and Dungmali. Linguistics of the Tibeto-Burman Area 14/1:73-92.

1992, Kiranti suffix copying (and possible origins of tense-aspect markers). Paper presented at the XXVth International Conference on Sino-Tibetan Languages and Linguistics, University of California at Berkeley, 14 October 1992.

Hagege, Claude, 1986, La langue palau, une curiosite typologique. Mtinchen: Wilhelm Fink Verlag.

Michailovsky, Boyd, 1989, Compte-rendu de A grammar of Limbu par George van Driem. Bulletin de la Societe de Linguistique de Paris 84/2:470-473.

Sprigg R. Keith, 1989, Review of A grammar of Limbu by George van Driem. Bulletin of the School of Oriental and African Studies 52/1:163-165.